BCI-MCP

An MCP server that streams live EEG brain-state metrics — focus, calm, attention, band powers, signal quality — from consumer/research EEG headsets (or a hardware-free synthetic source) into Claude and other MCP clients.

   
Type MCP server
Supplier enkhbold470 (community, MIT)
Availability Alpha — PyPI/npm bci-mcp v0.1.3, released 2026-06-24
Pricing Free / OSS (MIT)
Capabilities Read/Write — reads live/recorded EEG streams; can record sessions and run neurofeedback loops
Verified works · 2026-07-20
Security caution · 2026-07-20 — provenance matches supplier, MIT, no OSV advisories, single-maintainer alpha

How to install

  • Claude Code — direct MCP add (npm entrypoint): 
    claude mcp add bci-mcp -- npx -y bci-mcp
  • Claude Code — Python entrypoint (installs from PyPI on first run via uvx): 
    claude mcp add bci-mcp -- uvx bci-mcp serve
  • Claude Desktop — Settings → Developer → Edit Config, add: 
    {
  "mcpServers": {
    "bci-mcp": {
      "command": "npx",
      "args": ["-y", "bci-mcp"]
    }
  }
}
  • Manual PyPI install (for the bci-mcp CLI — stream, record, play, neurofeedback, dashboard): 
    pip install "bci-mcp[all]"

    Use extras [devices] (BrainFlow), [lsl] (pylsl), [edf] (recording playback), or [dashboard] to install only what you need.

This is a stdio server (built with FastMCP): the claude mcp add/Desktop config launch the process on demand — you do not keep it running in a separate terminal. Running uvx bci-mcp serve or npx -y bci-mcp by hand is only a one-shot check that it boots (Ctrl-C after it starts).

What it does

Exposes 13 tools: list_devices, connect, disconnect, get_brain_state, get_band_powers, get_signal_quality, get_metric_definitions, calibrate, record, start_neurofeedback, get_neurofeedback_score, mark_event, stream_summary.

Devices are addressed by URI scheme:

  • synthetic:// — built-in synthetic signal, no hardware required (lets you exercise the whole toolchain offline)
  • brainflow://cyton, brainflow://muse_s — OpenBCI Cyton/Ganglion, Muse 2/S via BrainFlow
  • lsl://<name> — any Lab Streaming Layer source (pylsl)
  • serial://<port> — generic serial device (pyserial)
  • playback://<file> — replay a previously recorded session

Primary use cases: neurofeedback experiments, live EEG band-power/brain-state monitoring in-conversation, recording and replaying EEG sessions, prototyping BCI workflows without hardware.

Notes

Community/independent project (not affiliated with any device vendor). Early-stage (v0.1.x, ~120 commits as of verification) — treat as Alpha. The synthetic:// source means Claude can drive the full tool surface with no headset attached, which is useful for evaluation; real-device use requires the matching hardware and BrainFlow/pylsl extras. A documentation-derived HTTP deployment mode also exists but the stdio path above is the supported Claude integration.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.